AMD FAB36 capacity reality check

I projected 40% market share (run rate) for AMD exiting 2006. But there are plenty of half-witted analysts who think AMD's capacity is only good for 25% market share maximum. Do they know to count their fingers? AMD is getting 21.4% unit share with one 200mm FAB!

Let's see how much FAB36 can do. First we compute the number of dual core Opteron die candidates for a 300mm wafer, using the Wafer program. Since dual core Opteron has a die area of 199 mm^2 at 90nm, at 65nm, the die area will be 100 mm^2. Using the Wafer program, we found that a 300mm wafer can produce at least 600 dual core Opteron dies.

We know AMD's yield is near 100%. Assume AMD is slow in ramping up FAB36, and they exit 2006 with only 8000 wspm, that's 4.8 million dual core opterons per month, or 58 million dual core Opterons out of FAB36 alone, annualy, or about 25% of the annual x86 units. Let's assume AMD gets unlucky with their 65nm process and yield will only be at 80%, we still get 46.4 million dual core Opterons per year. But remember, most of AMD CPUs are smaller than dual core Opterons, so the number will go up.

Now, as an exercise for the analysts, add the 21.4% unit share from the existing FAB30 (and don't forget this).

In another article, I concluded that INTEL's yield is only a fraction of AMD's yield.